COMMENT ⊗   VALID 00006 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "memo.pub[let,jmc]" source
C00006 00003	Outline of epistemology
C00007 00004	Experimental problems
C00009 00005	General notes
C00010 00006	.skip 1
C00011 ENDMK
C⊗;
.require "memo.pub[let,jmc]" source;
.turn off "{"
.cb EPISTEMOLOGICAL PROBLEMS OF COGNOLOGY


	A person's opinion on how long it will take to achieve human-level
artificial intelligence depends partly on how optimistic a personality he
has but even more on what problems he can see between the present state of
our science and this goal.  Unless he is convinced that the problems he
can see are all the important problems, a person can only give lower bounds on
the time required to overcome them.  Since 1958 (see McCarthy 1959), I
have seen the epistemological problems as major obstacles to artificial
intelligence, and I have always been disappointed at how few people work
seriously on them.  For example, when Erik Sandewall and I announced a
special issue of the AI Journal devoted to these problems, we received too
few relevant papers to justify the special issue.

	The neglect of epistemology seems related to
the fact that 95α% of the work in artificial intelligence is engineering
(trying to make a program with specified performance) and only a5α% is
science (studying a natural phenomenon).  I think it should be about
75-25.  For this reason and also to get a name more parallel to those of
other sciences, I suggest renaming our subject %2cognology%1 - %3the
science of intellectual processes and their relation to the problems they
solve%1.

	The %2epistemological%1 problems of cognology are clarified by trying
to separate them from the %2heuristic%1 problems.  %2Epistemology%1 studies
what kinds of facts are available for solving problems, and %2heuristics%1
studies how to use them to solve the problem.  Admittedly this separation
can only be partial, because there are facts about heuristics to be
represented and because the heuristic methods for dealing with facts
depend on what kinds of facts there are and how they are represented.
Nevertheless, we can identify many specifically epistemological problems
that lie between us and human-level artificial intelligence.

	Our first observation about the information available for
solving problems is that it is usually partial knowledge.

Outline of epistemology

Time, events, states, persistence

Objects, creation, persistence, destruction
Material

Space, location, motion

Causality and ability

Approximations, models, and correspondences

Persons - ability, goals

Information - knowledge, belief, desires and goals.
Experimental problems

	The traders buy and sell wheat by telephone.  There are at
least farmers and bakers, but perhaps also warehouse owners, wholesale
dealers and speculators.  We must represent such facts as, "If
I don't accept his offer now, he may buy what he wants from somebody
else before I can call him back".

	Missionaries and cannibals dialog with the heckler who
suggests that the boat may leak or there may be a bridge.

	More blocks.

	I need to book my flights with the travel agent, but I don't
need to know now the gate from which my plane will leave, but I
need to be able to find out in Bahrein even without speaking Arabic.

General notes

	Include a summary of concepts paper.

	The problem of refinable knowledge.
.skip 1
.begin verbatim
John McCarthy
Artificial Intelligence Laboratory
Computer Science Department
Stanford University
Stanford, California 94305

ARPANET: MCCARTHY@SU-AI
.end

.turn on "{"
%7This draft of

PUBbed at {time} on {date}.%1